Study design
This was a cross-sectional study. The study population included all the students aged between 10 and 18 years old at three academic grades in schools of Kerman, in 2020.
Sampling and participants
In this study, multi-stage sampling was used. At the first stage (stratified sampling), the selected schools were classified based on the students’ gender and the type of classification was based on the geographical districts of Kerman, the number of schools in each group, and the relative share of girls and boys in each area. At the second stage, based on the list of the students in each school, they were selected by simple random sampling using a random number table. Considering the number of grades, we performed the relativity distribution and then 3 grades were selected. As a result, 1100 students (560 boys and 540 girls) were selected from elementary schools, secondary schools, and high schools in Kerman, Iran, in 2020. Because some potential confounders may be present at different levels, middle school students (14-year-old subjects) were selected from junior high schools as the sampling framework.
Instruments
To investigate the prevalence of bullying behavior in the selected schools, the Iranian version of the Rezapour et al.’s questionnaire was used, which is based on the Olweus bullying questionnaire. According to the objectives of the study, only those questions measuring the prevalence of bullying and victimization, were asked. The students were also asked to indicate the number of times they had victimized or bullied in the past 2 or 3 months in the questionnaire.
Revised Questionnaire of Bully / Victim of Olweus (OB / VQ-R)
The initial Olweus questionnaire (Olweus, 1986) has 40 items and is scored on a 5-point scale, but its revised and abbreviated questionnaire (Olweus, 1998) has 28 items. The bully/ victim questionnaire mainly asks students if they have “bullied others in the past few months, or have been bullied by others.” Students are asked those questions that could best reflect their way of thinking or feeling during the past 2 or 3 months. Of note, the Olweus questionnaire (Olweus, 2002) is the most important and widely used sale of bullying worldwide, the validity and reliability of which have been confirmed in several studies (1, 23-25). Cronbach's alpha coefficient of 0.86 indicates the desirable and acceptable reliability of this questionnaire for assessing bullying among children and adolescents. In the present study, the reliability of this questionnaire was obtained by a test-retest methodology using Cronbach's alpha coefficient as 0.70 (26).
The questionnaire used in this study consists of 2 parts
- Demographic information of the participants
- Questions related to the victim (items 1 to 14) and bullying (items 15 to 18)
Part 1: This section included demographic information of the individuals and school level factors.
The level of students' economic and their social status were constructed using principal component analysis (PCA), which has been used in some other studies on bullying behaviors. In our study, several measurements were made of a combination of the available data on each family economic status (e.g., house characteristics) and parents' educational level and employment status. This combination was expected to better explain the level of students' economic and their social status and also to minimize the risk of observing multiple correlations in subsequent multivariate analyses. Economic level and social status indexes, in the form of the composite z score, were extracted as the first major components with the highest eigenvalue, and they were then used to explain the economic and social statuses of the included students adequately (61% of the variance in the measured socio-economic indicators). The score of the relevant economic and social status was classified into five categories (from low SES = 1 to high level SES = 5) and then used in the dimensional regression analysis. Moreover, the following variables were used in PCA: parents’ educational levels (illiterate = 1 to faculty members = 6), parents’ employment status (employed versus unemployed), home ownership (owner versus tenant), number of rooms per capita and having a private room (yes versus no), the amount of pocket money, and the number of family members.
Part 2: Item 1 has a general reference to victimization. Items 2, 8, and 9 are related to verbal victims. Items 3 and 5 are associated with communicative or emotional victims. Items 4, 6, and 7 are related to physical victims. Item 10 is related to cyber victims. Item 11 is related to other victims. Items 12, 13, and 14 describe the characteristics of a victim. Item 15 is a general reference to bullying. Items 16, 22, and 23 are related to verbal bullying. Items 17 and 19 are associated with communicative or emotional bullying. Items 18, 20, and 21 are related to physical bullying. Item 24 is related to cyber bullying. Item 25 is related to other cases of bullying. Finally, items 26, 27, and 28 describe the characteristics of bully subjects. Cutting points of 2 or 3 times a month are recommended as the most appropriate criterion for being a victim or a bully.
Data collection
Three teams experienced in bullying surveys were trained in a one-day workshop to review the objectives of this study and techniques of data collection. The team members were responsible for gathering the required information. Questions on school characteristics were answered by school principals, and items about bullying, victimization, and personal information were answered by the students. The students who were randomly selected (using a table of random numbers) from each class were then gathered in a single room and the questionnaires were distributed among them. The questionnaire was generally read by a facilitator to help the students in selecting the right option. Afterward, data were collected within 8 weeks from November 2019 to January 2020. Notably, all the data collection steps were monitored by the lead researcher.
To prevent bias caused by social undesirability, it was significantly controlled using a sealed ballot box. The students were assured that their answers would be cast in the ballot box without being handed over to the respondent. Conscious written consent was also obtained from all the respondents and they were free to leave the study at any desired time.
Data Analysis
Descriptive statistics were used to present the general characteristics of the selected schools and students (age, gender, students’ educational level, parents’ educational level, employment status of parents, etc.). Since the prevalence of bullying and victimization is a counting variable with high variability, in this study, the negative multiple regression model (backward method) was used to analyze the obtained data. Briefly, a mixed-effect negative binomial regression model with an independent variance-covariance structure was used as the final multivariate analysis in a generalization of the Poisson model.
To determine the final multivariate model, at first, we performed the univariate negative binomial regression analysis to measure the non-mediated effects of some factors on bullying behavior. Thereafter, as shown in Table 1, those variables with a significance level of less than 0.20 in the univariate analysis were simultaneously entered into negative binomial regression analysis with a multivariate mixed effect. In addition, each one of the main study criteria (e.g., perceived parental control) was adjusted for other variables (such as gender, SES, level of nutritional knowledge, media advertising, and perceived self-efficacy) in the model.
The minimum criteria measures of Akaike’s and Schwarz’s Bayesian information (namely AIC and BIC) were used to evaluate the good fit of the final model. Probability ratio test was also used as a criterion for applying the 2-level negative multiple regression model instead of using the negative multiple regression model. Subsequently, to better fit the model, the residual standard analysis was performed using the deviation and Pearson method. Finally, all the collected data were analyzed using SPSS v. 16.
Ethical considerations
The study permit was obtained from the Kerman Education Department. Informed consent was also obtained from the parents of all the students included. The needed data were collected using an anonymous self-report in the classroom by a trained researcher. The objectives and relevance of the study as well as the confidentiality of the students' responses were explained to them. The students were also informed about that the questionnaire completion was not mandatory. It is noteworthy that there was no time limit for completing the questionnaire and the average time to complete the questionnaires was about half an hour.
This project was approved with the code IR.KMU.REC.1397.568 on 2019/03/11 (date) by the ethics committee of Kerman University of Medical Sciences.
Findings
Table 1 shows both the percentage and frequency of variables of the individuals as well as the characteristics of the schools under study. In this study, 1100 students were enrolled and the response rate was 100%, among whom 540 (49.1%) subjects were boys and 560 (50.9%) subjects were girls. Of the studied schools, 877 (79.7%) schools were public and 223 (20.3%) were non-profit schools. Additionally, 420 students were studying in primary school, 374 students in middle school, and 603 students in high school. Moreover, among the studied schools, 510 had health promotion programs and in 590 schools, health promotion programs were not implemented. Other variables, including pocket money (in Iranian Rials), homeownership, having a private room, family members, living with parents, educational and employment statuses of parents, having health educator, principal’s experiences, having counselors, and number of bullying per week are presented in Table 1.
Using logistic regression model and considering the moderating variables, the results of this study show that the gender variable has a significant relationship with verbal, physical, and some other types of bullying, in a way that the chance of verbal, physical, and other types of bullying is 2.08, 1.89, and 2.22 times more among boys than girls, respectively. The results related to the variables modulating verbal, physical, and other types of bullying also show that those schools with health-promoting programs had a 1.45, 1.70, and 2.51 higher chance of developing verbal, physical, and other types of bullying than the schools without them. The moderating variables of both types of verbal and physical bullying have a significant relationship with the variable of studying level, in a way that the chances of verbal and physical bullying occurring in middle and high schools (0.69 and 0.65) and (0.22 and 0.48) were found to be lower than the chance on primary schools. The moderating variables of verbal and other types of bullying showed a significant relationship with the variable of living with other family members other than parents. Accordingly, the probability of verbal and other types of bullying in children living with other family members than the student living with both parents was estimated at 2.12 and 6.79 times higher, respectively. Additionally, the moderating variable of physical bullying was indicated to have a significant relationship with the variables of living with one of the parents and living with other family members. In this regard, the probability of physical bullying in living with one parent and other family members was estimated at 2.66 and 4.53 times higher than living with both parents, respectively. The results show that no significant relationship exists between moderator and independent variables such as type of school, presence of counselor, principal's experience, number of students, and the factors related to economic and social statuses (including having private room, the amount of pocket money, parents' educational levels, room per person, employment status, number of family members, and monthly family income) (P value> 0.05) (Table 2).
Using the multi logistic regression model, the results of this study indicate that the moderator variable of bullied and bully persons have a significant relationship with the variable of gender, so that the probability of bullying and being bullied is 2.36 times higher among boys than girls. With respect to the variable of schools providing health-promoting programs, a significant relationship was found with the moderator variable of emotional bullying, indicating that emotional bullying is 1.80 times more likely to occur in the schools with health-promoting programs than the schools without them. In the current study, the results demonstrate that the modulator variables of emotional bullying and being bullied have a significant relationship with the variable of the presence of counselors in schools, in a way that the chances of emotional bullying and being bullied were calculated to be 1.65 and 2.92 times higher in the schools with counselors, respectively. A significant relationship was also observed between the moderator variable of emotional bullying and the variable of educational level, in a way that the probability of the occurrence of emotional bullying in the middle and high schools is 0.35 and 0.44 times lower than the elementary school, respectively. As well, the chance of bullying and being bullied in the middle school is 0.37 less than this chance in elementary school. The results show that the moderating variable of emotional bullying has a significant relationship with the variable of living with parents. In this regard, the probability of emotional bullying in students living with one parent or living with other family members is 2.02 and 2.95 times higher than that of the students living with both parents. There is no significant relationship between the variables regulating emotional bullying and being bullied and the variables of school type, number of students, school’s principal’s experience, and economic and social statuses (P value> 0.05) (Table 3).
The results obtained using the logistic regression model showed that all the moderating variables of the perpetrator of verbal, physical, and cyber bullying were significantly related to the following variables: gender, health-promoting schools, educational levels, living with other family members, and low economic level and social status (P value <0.05). With respect to the variable of gender, the chances of boys committing verbal, physical, and cyber bullying were found to be 3.27, 8.59, and 15.72 times higher than girls, respectively. In terms of the variable of the schools with health-promoting programs, the chance of verbal, physical, and cyber bullying was estimated as 1.71, 2.18, and 5.32 times higher than in the schools without health promoting programs, respectively. Moreover, the probability of verbal and physical bullying high school students is 2.19 and 2.48 times higher than primary school students, respectively. As well, the probability of cyber bullying by the students studying in middle and high schools is 0.09 less and 5.99 higher than elementary school, respectively. The results show that the chance of committing cyber bullying by students living with one of their family members is 10.70 times higher than that of students living with both parents. Furthermore, the probability of the moderating variable of physical bullying by the students with low economic and social statuses was found to be 0.15 less than the students with higher economic and social statuses. According to the obtained results, no significant relationship was observed between moderating variables and other variables such as type of schools, presence of counselor, experience of the schools’ principals, and number of students (P value> 0.05) (table 4).
The results of the multi logistic regression model showed that a significant relationship exists between the moderator variable of emotional bullying and the variable of gender, and also among the moderator variable of bullying and the variables of gender, middle school level, and living with other family members (P-value <0.05). In this regard, in terms of the variable of gender, the chance of emotional bullying was calculated to be 3.40 and 6.96 times higher in boys compared to girls. As well, based on educational level, the probability of bullying is 0.28 times lower among middle school students compared to elementary school students. In regard with the variable of living with other family members other than the parents, the probability of bullying among the students was observed to be 3.92 times higher than that of the students living with their parents. There was no significant relationship amongst the moderating variables of emotional bullying and type of school, schools with health-promoting programs, presence of counselor, the schools’ principal experience, number of students, and their economic and social statuses (P-value> 0.05). (Table 5)